Research map / 3D vision / 3D generation
OmniNWM: Unifying the State-Action-Reward Triad for Closed-Loop Panoramic Driving Navigation World Models
Bohan Li, Zhuang Ma, Dalong Du, Baorui Peng, Zhujin Liang, Zhenqiang Liu, Xianda Guo, Chao Ma, Yueming Jin, Zheng Zhu, HAO ZHAO, Wenjun Zeng and 1 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add canonical panoramic ray-map normalization (ablation) | OmniNWM · nuScenes camera control | TransErr (m): 9.75 -> 1.18 | Improved |
| add canonical panoramic ray-map normalization (ablation) | OmniNWM · nuScenes camera control | RotErr (rad): 1.71 -> 0.16 | Improved |
| add structured panoramic forcing instead of pure autoregression (ablation) | OmniNWM · nuScenes 201-frame generation | FVD201: 386.72 -> 25.22 | Improved |
| use structured noise instead of standard scheduled sampling (ablation) | OmniNWM · nuScenes 201-frame generation | FVD201: 178.65 -> 25.22 | Improved |
| add unstructured scheduled sampling noise to pure autoregression (ablation) | OmniNWM · nuScenes 201-frame generation | FVD201: 386.72 -> 178.65 | Improved |
| add generated semantics and depth as occupancy inputs (ablation) | OmniNWM occupancy generator · nuScenes-Occupancy | mIoU: 3.0 and 2.7 mIoU | Improved |
| OmniNWM vs UniScene on video quality | OmniNWM · nuScenes val | FVD: 71.94 -> 23.63 | Improved |
| OmniNWM vs DiST-4D zero-shot on nuPlan | OmniNWM · nuPlan zero-shot | TransErr (m): 7.12 -> 1.65 | Improved |
| OmniNWM vs Hi-SOP on occupancy | OmniNWM · nuScenes-Occupancy | mIoU: 16.4 -> 19.8 | Improved |
| OmniNWM vs M2Depth on panoramic depth | OmniNWM · nuScenes val depth | Abs. Rel.: 0.26 -> 0.23 | Improved |
Later papers that compared with this method
1 later accepted paper reports OmniNWM as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Doe-2: 3D Representation World Model for Unified Driving Scene Forecasting (ECCV 2026 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- M2Depth — M^2Depth: Self-supervised Two-Frame Multi-camera Metric Depth Estimation (ECCV 2024) · better on 2 of 2 shared results
- OccScene — OccScene: Semantic Occupancy-based Cross-task Mutual Learning for 3D Scene Generation (TPAMI 2025) · better on 1 of 1 shared result
- OccGen — OccGen: Generative Multi-modal 3D Occupancy Prediction for Autonomous Driving (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Driving (ECCV 2026 · similarity 0.89)
- Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving (AAAI 2025 · similarity 0.87)
- PanoWorld: Geometry-Consistent Panoramic Video World Modeling (NeurIPS 2026 · similarity 0.86)
- Navigation World Models (CVPR 2025 · similarity 0.86)
- Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllability (NeurIPS 2024 · similarity 0.86)
- DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
